Renewing Philosophy 9780674042384

Citation preview

RENEWING

PHILOSOPHY

RENEWING PH I LOSOPHY

Hilary Putnam

HARVARD UNIVERSITY PRESS

Cambridge, Massachusetts London, England

Copyright © 1992 by the President and Fellows of Harvard College All rights reserved Printed in the United States of America This book has been digitally reprinted. The content remains identical to that of previous printings.

Library of Congress Cataloging-in-Publication Data Putnam, Hilary. Renewing philosophy / Hilary Putnam. p. em. Includes bibliographical references and index. ISBN 0-674-76093-X (alk. paper) (cloth) ISBN 0-674-76094-8 (pbk.) 1. Philosophy. 2. Philosophy and science. I. Title. B29.P88 1992 92-10854 100-dc20 CIP

For Erika, Samuel, Joshua, and Polly (Max)

Contents

Preface

IX

I. The Project of Artificial Intelligence

2. Does Evolution Explain Representation?

19

3. A Theory of Reference

35

4. Materialism and Relativism

60

5. Bernard Williams and the Absolute Conception of the World

80

6. Irrealism and Deconstruction

108

7. Wittgenstein on Religious Belief

134

8. Wittgenstein on Reference and Relativism

158

9. A Reconsideration of Deweyan Democracy

180

Notes

203

Index

227

vii

Preface

The present book grew out of the Gifford Lectures, which I delivered at the University of St. Andrews in the Fall of 1990, and, with one exception, its chapters are quite close to the lectures as given. (Chapter 5 has been very substantially rewritten. In addition, there was an opening lecture in which, perhaps perversely, I chose to deal with the present situation in quantum mechanics and its philosophical significance, which I decided did not really belong with the others.) At first blush, the topics with which the lectures dealt may seem to have little relation to one another: I spoke of reference and realism and religion and even of the foundations of democratic politics. Yet my choice of these topics was not an arbitrary one. I was guided, of course, by my own past areas of concern, since it would have been foolish to lecture on topics on which I had not done serious thinking and writing in the past, but beyond that I was guided by a conviction that the present situation in philosophy is one that calls for a revitalization, a renewal, of the subject. Thus this book, in addition to addressing several topics individually, offers a diagnosis of the present situation in philosophy as a whole and suggests the directions in which we might look for such a renewal. That suggestion does not take the form of a manifesto, however, but rather takes the form of a series of reflections on various philosophical ideas. Analytic philosophy has become increasingly dominated by ix

PREFACE

x the idea that science, and only science, describes the world as it is in itself, independent of perspective. To be sure, there are within analytic philosophy important figures who combat this scientism: one has only to mention Peter Strawson, or Saul Kripke, or John McDowell, or Michael Dummett. Nevertheless, the idea that science leaves no room for an independent philosophical enterprise has reached the point at which leading practitioners sometimes suggest that all that is left for philosophy is to try to anticipate what the presumed scientific solutions to all metaphysical problems will eventually look like. (This is accompanied by the weird belief that one can anticipate that on the basis of present-day science!) The first three chapters in this volume are concerned to show that there is extremely little to this idea. I begin with a look at some of the ways in which philosophers have suggested that modern science explains the link between language and the world. The first chapter discusses the decidedly premature enthusiasm that some philosophers feel for HArtificial Intelligence". The second chapter takes on the idea that evolutionary theory is the key to the phenomenon of representation, while the third chapter subjects to close scrutiny a contemporary philosopher's claim that one can define reference in terms of causality. I try to show that these ideas lack scientific and philosophical substance, while gaining prestige from the general philosophical climate of deference to the supposed metaphysical significance of science. Perhaps the most impressive case for the view that one should look to present-day science, and especially to physics, for at least a very good sketch of an adequate metaphysics has been made by the British philosopher Bernard Williams, and after a chapter which deals with some of the problems faced by both relativistic and materialistic metaphysicians, I devote a chapter to a close examination of his views. Not all present-day philosophers are overawed by science,

PREFACE

xi

however, and some of the philosophers who are not-philosophers like Derrida, or, in the English-speaking world, Nelson Goodman or Richard Rorty-have reacted to the difficulty of making sense of our cognitive relation to the world by denying that we do have a cognitive relation to extralinguistic reality. In my sixth chapter, I criticize these thinkers for throwing away the baby with the bathwater. In the seventh and eighth chapters, I examine Wittgenstein's "Lectures on Religious Belief", arguing that those lectures demonstrate how a philosopher can lead us to see our various forms of life differently without being either scientistic or irresponsibly metaphysical, while in the concluding chapter I try to show how John Dewey's political philosophy exhibits the same possibility in a very different way. The two months that I spent at St. Andrews giving these lectures were a sheer delight, and I profited more than I can say from the companionship and the philosophical conversation of the remarkable group of brilliant and dedicated philosophers there, particularly Peter Clark, Bob Hale, John Haldane, Stephen Read, Leslie Stevenson, John Skorupski, and Crispin Wright. As always in recent years, many of the ideas in these chapters were first tried out in conversation with Jim Conant, and Chapter 5, in particular, owes a great deal to those conversations. Chapter 9 first appeared, in a slightly different form, in Southern California Law Review 63 (1990): 1671-97, and is reprinted here with that journal's permission. I am also grateful to Bengt Molander of the University of Uppsala and to BenAmi Sharfstein of the University of Tel Aviv, both of whom read earlier versions and made valuable suggestions. At a very late stage, excellent suggestions were also made by the referees for the Harvard University Press, not all of which I could take up without changing the character of the work, but some of which I have responded to, and some of which will show their effect in my future writing. The most valuable suggestions of

PREFACE

xii

all were made by Ruth Anna Putnam, who provided not only the affection and support which mean so much, but whose close reading and fine criticism certainly made this a much better book.

RENEWING

PHILOSOPHY

1 The Project of Artificial Intelligence

Traditionally Gifford Lectures have dealt with questions connected with religion. In recent years, although reference to religion has never been wholly absent, they have sometimes been given by scientists and philosophers of science, and have dealt with the latest knowledge in cosmology, elementary particle physics, and so on. No doubt the change reflects a change in the culture, and particularly in the philosophical culture. But these facts about the Gifford Lectures-their historical concern with religion and their more recent concern with scienceboth speak to me. As a practicing Jew, I am someone for whom the religious dimension of life has become increasingly important, although it is not a dimension that I know how to philosophize about except by indirection; and the study of science has loomed large in my life. In fact, when I first began to teach philosophy, back in the early 1950s, I thought of myself as a philosopher of science (although I. included philosophy of language and philosophy of mind in my generous interpretation of the phrase ((philosophy of science"). Those who know my writings from that period may wonder how I reconciled my religious streak, which existed to some extent even back then, and my general scientific materialist worldview at that time. The answer is that I didn't reconcile them. I was a thoroughgoing atheist, and I was a believer. I simply kept these two parts of myself separate.

RENEWING PHILOSOPHY

2 In the main, however, it was the scientific materialist that was dominant in me in the fifties and sixties. I believed that everything there is can be explained and described by a single theory. Of course we shall never know that theory in detail, and even about the general principles we shall always be somewhat in error. But I believed that we can see in present-day science what the general outlines of such a theory must look like. In particular, I believed that the best metaphysics is physics, or, more precisely, that the best metaphysics is what the positivists called "unified science", science pictured as based on and unified by the application of the laws of fundamental physics. In our time, Bernard Williams has claimed that we have at least a sketch of an "absolute conception of the world" in present-day physics. 1 Many analytic philosophers today subscribe to such a view, and for a philosopher who subscribes to it the task of philosophy becomes largely one of commenting on and speculating about the progress of science, especially as it bears or seems to bear on the various traditional problems of philosophy. When I was young, a very different conception of philosophy was represented by the work of John Dewey. Dewey held that the idea of a single theory that explains everything has been a disaster in the history of philosophy. Science itself, Dewey once pointed out, has never consisted of a single unified theory, nor have the various theories which existed at anyone time ever been wholly consistent. While we should not stop trying to make our theories consistent-Dewey did not regard inconsistency as a virtue-in philosophy we should abandon the dream of a single absolute conception of the world, he thought. Instead of seeking a final theory-whether it calls itself an "absolute conception of the world" or not-that would explain everything, we should see philosophy as a reflection on how human beings can resolve the various sorts of "problematical situations" that

THE PROJECT OF ARTIFICIAL INTELLIGENCE

3 they encounter, whether in science, in ethics, in politics, in education, or wherever. My own philosophical evolution has been from a view like Bernard Williams' to a view much more like John Dewey's. In this book I want to explain and, to the extent possible in the space available, to justify this change in my philosophical attitude. In the first three chapters, I begin with a look at some of the ways in which philosophers have suggested that modern cognitive science explains the the link between language and the world. This chapter deals with Artificial Intelligence. Chapter 2 will discuss the idea that evolutionary theory is the key to the mysteries of intentionality (i. e., of truth and reference), while Chapter 3 will discuss the claim made by the philosopher Jerry Fodor that one can define reference in terms of causal/counterfactual notions. In particular, I want to suggest that we can and should accept the idea that cognitive psychology does not simply reduce to brain science cum computer science, in the way that so many people (including most practitioners of "cognitive science") expect it to. I just spoke of a particular picture of what the scientific worldview is, the view that science ultimately reduces to physics, or at least is unified by the world picture of physics. The idea of the mind as a sort of "reckoning machine" goes back to the birth of that "scientific worldview" in the seventeenth and eighteenth centuries. For example, Hobbes suggested that thinking is appropriately called "reckoning", because it really is a manipulation of signs according to rules (analogous to calculating rules), and La Mettrie scandalized his time with the claim that man is just a machine (L'Homme Machine). 2 These ideas were, not surprisingly, associated with materialism. And the question which anyone who touches on the topic of Artificial Intelligence is asked again and again is "Do you think that a computing machine could have intelligence, conscious-

RENEWING PHILOSOPHY

4

ness, and so on, in the way that human beings do?" Sometimes the question is meant as "could it in principle" and sometimes as "could it really, in practice" (to my mind, the far more interesting question). The story of the computer, and of Alan Turing's role in the conception of the modern computer, has often been told. In the thirties, Turing formulated the notion of computability3 in terms which connect directly with computers (which had not yet been invented). In fact, the modern digital computer is a realization of the idea of a "universal Turing machine". A couple of decades later materialists like my former self came to claim that "the mind is a Turing machine". It is interesting to ask why this seemed so evident to me (and still seems evident to many philosophers of mind). If the whole human body is a physical system obeying the laws of Newtonian physics, and if any such system, up to and including the whole physical universe, is at least metaphorically a machine, then the whole human body is at least metaphorically a machine. And materialists believe that a human being is just a living human body. So, as long as they assume that quantum mechanics cannot be relevant to the philosophy of mind (as I did when I made this suggestion), 4 materialists are committed to the view that a human being is-at least metaphorically-a machine. It is understandable that the notion of a Turing machine might be seen as just a way of making this materialist idea precise. Understandable, but hardly well thought out. The problem is the following: a "machine" in the sense of a physical system obeying the laws of Newtonian physics need not be a Turing machine. (In defense of my former views, I should say that this was not known in the early 1960s when I proposed my so-called functionalist account of mind. ) For a Turing machine can compute a function only if that function

THE PROJECT OF ARTIFICIAL INTELLIGENCE

5 belongs to a certain class of functions, the so-called general recursive functions. But it has been proved that there exist possible physical systems whose time evolution is not describable by a recursive function, even when the initial condition of the system is so describable. (The wave equation of classical physics has been shown to give rise to examples.) In less technical language, what this means is that there exist physically possible analogue devices which can "compute" non-recursive functions. 5 Even if such devices cannot actually be prepared by a physicist (and Georg Kreisel has pointed out that no theorem has been proved excluding the preparation of such a device),6 it does not follow that they do not occur in nature. Moreover, there is no reason at all why the real numbers describing the condition at a specified time of a naturally occurring physical system should be "recursive". So, for more than one reason, a naturally occurring physical system might well have a trajectory which "computed" a non-recursive function. You may wonder, then, why I assumed that a human being , could be, at least as a reasonable idealization, regarded as a Turing machine. One reason was that the following bit of reasoning occurred to me. A human being cannot live forever. A human being is finite in space and time. And the words and actions-the "outputs", in computer jargon-of a human being, insofar as they are perceivable by the unaided senses of other human beings (and we might plausibly assume that this is the level of accuracy aimed at in cognitive psychology) can be described by physical parameters which are specified to only a certain macroscopic level of accuracy. But this means that the "outputs" can be predicted during the finite time the human lives by a sufficiently good approximation to the actual continuous trajectory, and such a "sufficiently good approximation" can be a recursive function. (Any function can be approximated to any fixed level of accuracy by a recursive function over any

RENEWING PHILOSOPHY

6

finite time interval.) Since we may assume that the possible values of the boundary parameters are also restricted to a finite range, a finite set of such rec~rsive functions will give the behavior of the human being under all possible conditions in the specified range to the desired accuracy. (Since the laws of motion are continuous, the boundary conditions need only to be known to within some appropriate ~ in order to predict the trajectory of the system to within the specified accuracy.) But if that is the case, the "outputs"-what the human says and does-ean be predicted by a Turing machine. (In fact, the Turing machine only has to compute the values of whichever recursive function in the finite set corresponds to the values that the boundary conditions have taken on), and such a Turing machine could, in principle, simulate the behavior in question as well as predict it. This argument proves too much and too little, however. On the one hand, it proves that every physical system whose behavior we want to know only up to some specified level of accuracy and whose ((lifetime" is finite can be simulated by an automaton! But it does not prove that such a simulation is in any sense a perspicuous representation of the behavior of the system. When an airplane is flying through the air at less than supersonic speeds, it is perspicuous to represent the air as a continuous liquid, and not as an automaton. On the other hand it proves too little from the point of view of those who want to say that the real value of computational models is that they show what our "competence" is in idealization from such limitations as the finiteness of our memory or our lifetimes. According to such thinkers,7 if we were able to live forever, and

were allowed access to a potentially infinite memory storage, still all our linguistic behavior could be simulated by an automaton. We are best "idealized" as Turing machines, such thinkers

THE PROJECT OF ARTIFICIAL INTELLIGENCE

7 say, when what is at stake is not our actual uperformance" but our ucompetence". Since the proof of the little theorem I just demonstrated depended essentially on assuming that we do not live forever and on assuming that the boundary conditions have a finite range (which excludes a potentially infinite external memory), it offers no comfort to such a point of view. Again, it might be said that any non-recursivities either in our initial conditions or in our space-time trajectories could not be reliably detected and hence would have no ucognitive" significance. But it is one thing to claim that the particular non-recursive function a human might compute if the human (under a certain idealization) were allowed to live forever has no cognitive significance, and another to say that the whole infinite trajectory can therefore be approximated by a Turing machine. Needless to say, what follows the Utherefore" in this last sentence does not follow logically from the antecedent! (Recall how in the uchaos" phenomena small perturbations become magnified in the course of time. ) In sum, it does not seem that there is any principled reason why we must be perspicuously representable as Turing machines, even assuming the truth of materialism. (Or any reason why we must be representable in this way at all-even nonperspicuously-under the idealization that we live forever and have potentially infinite external memories). That is all I shall say about the question whether we are (or can be represented as) Turing machines Uin principle". On the other hand, the interesting question is precisely whether we are perspicuously representable as Turing machines, even if there are no a priori answers to be had to this question. And this is something that can be found out only by seeing if we can "simulate" human intelligence in practice. Accordingly, it is to this question that I now turn.

RENEWING PHILOSOPHY

8

Induction and Artificial Intelligence A central part of human intelligence is the ability to make inductive inferences, that is, to learn from experience. In the case of deductive logic, we have discovered a set of rules which satisfactorily formalize valid inference. In the case of inductive logic this has not so far proved possible, and it is worthwhile pausing to ask why. In the first place, it is not clear just how large the scope of inductive logic is supposed to be. Some writers consider the Uhypothetico-deductive method"-that is, the inference from the success of a theory's predictions to the acceptability of the theory-the most important part of inductive logic, while others regard it as already belonging to a different subject. Of course, if by induction we mean uany method of valid inference which is not deductive", then the scope of the topic of inductive logic will be simply enormous. If the success of a large number of predictions-say, a thousand, or ten thousand-which are not themselves consequences of the auxiliary hypotheses alone always confirmed a theory, then the hypothetico-deductive inference, at least, would be easy to formalize. But problems arise at once. Some theories are accepted when the number of confirmed predictions is still very small-this was the case with the general theory of relativity, for example. To take care of such cases, we postulate that it is not only the number of confirmed predictions that matters, but also the elegance or simplicity of the theory: but can such quasi-aesthetic notions as "elegance" and usimplicity" really be formalized? Formal measures have indeed been proposed, but it cannot be said that they shed any light on real-life scientific inference. Moreover, a confirmed theory sometimes fits badly with background knowledge; in some cases, we conclude the theory cannot be true, while in others we conclude that the

THE PROJECT OF ARTIFICIAL INTELLIGENCE

9 background knowledge should be modified; again, apart from imprecise talk about "simplicity", it is hard to say what determines whether it is better, in a concrete case, to preserve background knowledge or to modify it. And even a theory which leads to a vast number of successful predictions may not be accepted if someone points out that a much simpler theory would lead to those predictions as well. In view of these difficulties, some students of inductive logic would confine the scope of the subject to simpler inferencestypically, to the inference from the statistics in a sample drawn from a population to the statistics in the population. When the population consists of objects which exist at different times, including future times, the present sample is never going to be a random selection from the whole population, however; so the key case is this: I have a sample which is a random selection from the members of a population which exist now, here (on Earth, in Scotland, in the particular place where I have been able to gather samples, or wherever); what can I conclude about the properties of future members of the population (and of members in other places)? If the sample is a sample of uranium atoms, and the future members are in the near as opposed to the cosmological future, then we are prepared to believe that the future members will resemble present members, on the average. If the sample is a sample of people, and the future members of the population are not in the very near future, then we are less likely to make this assumption, at least if culturally variable traits are in question. Here we are guided by background knowledge, of course. This has suggested to some inquirers that perhaps all there is to induction is the skilful use of background knowledge-we just "bootstrap" our way from what we know to additional knowledge. But then the cases in which we don't have much background knowledge at all, as well as the exceptional cases

RENEWING PHILOSOPHY

10 in which what we have to do is question background knowledge, assume great importance; and here, as just remarked, no one has much to say beyond vague talk about usimplicity". The problem of induction is not by any means the only problem confronting anyone who seriously intends to simulate human intelligence. Induction, indeed all cognition, presupposes the ability to recognize similarities among things; but similarities are by no means just constancies of the physical stimulus, or simple patterns in the input to the sense organs. For this reason, the success certain computer programs have had in detecting patterns (e. g., the shapes of letters of the alphabet) does not solve the usimilarity" problem in the form in which it confronts someone learning a natural language. What makes knives similar, for example, is not that they all look alike (they don't), but that they are all manufactured to cut or stab;8 any system that can recognize knives as relevantly similar needs to be able to attribute purposes to agents. Humans have no difficulty in doing this; but it is not clear that we do this by unaided induction; we may well have a uhard-wired-in" ability to "put ourselves in the shoes" of other people which enables us to attribute to them any purposes we are capable of attributing to ourselves-an ability that Evolution the Tinker found it convenient to endow us with, and which helps us to know which of the infinitely many possible inductions we might consider is likely to be successful. Again, to recognize that a chihuahua and a Great Dane are similar in the sense of belonging to the same species requires the ability to realize that, appearances notwithstanding, 9 chihuahuas can impregnate Great Danes and produce fertile offspring. Thinking in terms of potential for mating and potential for reproduction is natural for us; but it need not be natural for an artificial intelligenceunless we deliberately simulate this human propensity when we

THE PROJECT OF ARTIFICIAL INTELLIGENCE

11 construct the artificial intelligence. Such examples can be multiplied indefinitely. Similarities expressed by adjectives and verbs rather than nouns can be even more complex. A non-human "intelligence" might know what white is on a color chart, for example, without being able to see why pinko-grey humans are called "white", and it might know what it is to open a door without being able to understand why we speak of opening a border (or opening trade). There are many words (as Wittgenstein pointed out) that apply to things that have only a "family resemblance" to one another; there need not be one thing all X's have in common. For example, we speak of the Canaanite tribal chiefs mentioned in the bible as kings although their kingdoms were probably little more than villages, and we speak of George VI (who did not literally rule England at all) as a king; and there are even cases in history in which "the kingship was not hereditary", we say. Similarly (Wittgenstein's example), there is no property all games have in common which distinguishes them from all the activities which are not games. The notional task of artificial intelligence is to simulate intelligence, not to duplicate it. So, perhaps one might finesse the problems I just mentioned by constructing a system that reasoned in an ideal languagelO-one in which words did not change their extensions in a context-dependent way (a sheet of typing paper might be "whitel" and a human being might be "whitez", in such a language, where whitel is color-chart white, and whitez is pinko-grey). Perhaps all "family resemblance" words would have to be barred from such a language. (How much of a vocabulary would be left?) But my budget of difficulties is not yet finished. Because the project of symbolic inductive logic appeared to run out of steam after Carnap, the thinking among philosophers

RENEWING PHILOSOPHY

12 of science has run in the direction of talking about so-called bootstrapping methods-that is, methods which attribute a great deal to background knowledge. It is instructive to see why this has happened, and also to realize how unsatisfactory such an approach is if our aim is to simulate intelligence. One huge problem might be described as the existence of conflicting inductions. To use an example from Nelson Goodman:}} as far as I know, no one who has ever entered Emerson Hall in Harvard University has been able to speak Inuit (Eskimo). Thinking formalistically, this suggests the induction that if any person X enters Emerson Hall, then X does not speak Inuit. Let Ukuk be an Eskimo in Alaska who speaks Inuit. Shall I predict that if Ukuk enters Emerson Hall, then Ukuk will no longer be able to speak Inuit? Obviously not, but what is wrong with this induction? Goodman answers that what is wrong with the inference is that it conflicts with the "better entrenched" inductively supported law that people do not lose their ability to speak a language upon entering a new place. But how am I supposed to know that this law does have more confirming instances than the regularity that no one who enters Emerson Hall speaks Inuit? Background knowledge again? As a matter of fact, I don't believe that as a child I had any idea how often either of the conflicting regularities in the example (conflicting in that one of them must fail if Ukuk enters Emerson Hall) had been confirmed, but I would still have known enough not to make the Usilly" induction that Ukuk would stop being able to speak Inuit if he entered a building (or a country) where no one had spoken Inuit. Again it is not clear that the knowledge that one doesn't lose a language just like that is really the product of induction-perhaps this is something we have an innate propensity to believe or, if that seems unreasonable, something that we have an innate pro-

THE PROJECT OF ARTIFICIAL INTELLIGENCE

13 pensity to conclude on the basis of only a little experience. The question that won't go away is how much what we call "intel-

ligence" presupposes the rest of human nature. Moreover, if what matters really is "entrenchment" (that is, number and variety of confirming instances), and if the information that the universal statement "one doesn't lose one's ability to speak a language upon entering a new place" is better entrenched than the universal statement "no one who enters Emerson Hall speaks Inuit" is part of my background knowledge, it isn't clear how it got there. Perhaps this information is implicit in the way people speak about linguistic abilities; but then one is faced with the question of how one "decodes" the implicit information conveyed by the utterances one hears. The problem of conflicting inductions is a ubiquitous one even if one restricts attention to the simplest inductive inferences. If the solution is really just to give the system more background knowledge, then what are the implications for Artificial Intelligence? It is not easy to say, because Artificial Intelligence as we know it doesn't really try to simulate intelligence at all; simulating intelligence is only its notional activity, while its real activity is just writing clever programs for a variety of tasks. This is an important and useful activity, although, of course, it does not sound as exciting as "simulating human intelligence" or "producing artificial intelligence". But if Artificial Intelligence existed as a real rather than a notional research activity, there would be two alternative strategies its practitioners could follow in the face of the problem of background knowledge. (I) One could simply try to program into the machine all of the information a sophisticated human inductive judge has (including the tacit information). At the least it would require generations of researchers to formalize this information (probably it could not be done at all, because of the sheer quantity

RENEWING PHILOSOPHY

14 of information involved); and it is not clear that the result would be more than a gigantic "expert system". No one would find this very exciting; and such an "intelligence" would be dreadfully unimaginative, unable to realize that in many cases it is precisely background knowledge that needs to be given up. (2) One could undertake the more exciting and ambitious task of constructing a device that could learn the background knowledge by interacting with human beings, as a child learns a language and all the cultural information, explicit and tacit, that comes with growing up in a human community.

The Natural Language Problem The second alternative is certainly the project that deserves the name of Artificial Intelligence. But consider the problems: to figure out the information implicit in the things people say, the machine must simulate "understanding" a human language. Thus the idea mentioned above, of sticking to an artificial "ideal language" and ignoring the complexities of natural language, has to be abandoned if this strategy is adopted; abandoned because the cost is too high. Too much of the information the machine would need is retrievable only via natural language proceSSIng. But the natural language problem presents many of the same difficulties all over again. Some thinkers-Chomsky and his school-believe that a "template" for natural language, including the semantic or conceptual aspects, is innate-hard-wiredin by Evolution the Tinker. Although this view is taken to extremes by Fodor, who holds that there is an innate language of thought, with primitives adequate for the expression of all concepts that humans are able to learn to express in a natural language, Chomsky himself has hesitated to go this far: it seems that what he is committed to is the existence of a large number

THE PROJECT OF ARTIFICIAL INTELLIGENCE

15 of innate conceptual abilities which give us a propensity to form certain concepts and not others. (In conversation, he has suggested that the difference between postulating innate concepts and innate abilities is not important if the postulated abilities are sufficiently structured.) At the opposite extreme, there is the view of classical behaviorism, which sought to explain language learning as a special case of the application of general rules for acquiring "habits"-i. e., as just one more bundle of inductions. (An in-between position is, of course, possible: why should language learning not depend partly on special-purpose heuristics and partly on general learning strategies-both developed by evolution?) The view that language learning is not really learning, but rather the maturation of an innate ability in a particular environment (somewhat like the acquisition of a bird call by a species of bird that has to hear the call from an adult bird of the species to acquire it, but which also has an innate propensity to acquire that sort of call) leads, in its extreme form, to pessimism about the likelihood that human use of natural language can be successfully simulated on a computer-which is why Chomsky is pessimistic about projects for natural language computer processing, although he shares the computer model of the brain, or at least of the "language organ", with AI researchers. Notice that this pessimistic view of language learning parallels the pessimistic view that induction is not a single ability, but rather a manifestation of a complex human nature whose computer simulation would require a vast system of subroutines-so vast that generations of researchers would be required to formalize even a small part of the system. Similarly, the optimistic view that there is an algorithm (of manageable size) for inductive logic is paralleled by the optimistic view of language learning: that there is a more or less topic-neutral heuristic for learning, and that this heuristic suffices (without the

RENEWING PHILOSOPHY

16 aid of an unmanageably large stock of hard-wi red-in background knowledge, or topic-specific conceptual abilities) for the learning of one's natural language, as well as for the making of inductive inferences in general. Perhaps the optimistic view is right; but I do not see anyone on the scene, in either Artificial Intelligence or inductive logic, who has any interesting ideas as to how the topic-neutral learning strategy works.

The Mind as Chaos When I published a paper with these arguments,12 the American philosopher Daniel Dennett 13 characterized my view as "the mind as chaos." This is an interesting charge. Up to now I have been discussing the prospects of simulating human intelligence, not the prospects of finding informative models of the way the brain works. Dennett is connecting the two tasks: in effect, he is claiming that pessimism about the success of AI in simulating human intelligence amounts to pessimism about the possibility of describing the functioning of the brain. Hidden in this charge is a variant of Pascal's wager: you have nothing to lose if you assume AI will succeed and you are wrong, but if you assume AI will not succeed, you will lose the only chance there is to describe the brain. But what connection is there between simulating intelligence and describing the brain? Even if the computer model of the brain is correct, it does not at all follow that AI will succeed. As mentioned above, Noam Chomsky believes the computer model is correct, but he does not expect AI to succeed. Language-using, he once put it to me in conversation, is not a separable ability of human beings: you can simulate baseball-throwing without simulating total human intellectual capacity, but you cannot simulate

THE PROJECT OF ARTIFICIAL INTELLIGENCE

17 language-using-even language-using in a fixed context, such as going to the store and buying some milk, without simulating total human intellectual capacity. Yet Chomsky does not despair of understanding the brain; we can understand the weather without being able to predict it any better than we could before, and we may understand the brain, as a hierarchically structured system of computational systems ("modules"), without being able to describe all of them and all of their interactions well enough to predict or even simulate the brain's activities. Another example which makes the same point is the current interest in computer models of the brain which do not assume that the brain computes using "representations" and rules for manipulating those representations in the style of a logical calculus. 14 Perhaps the most interesting of these is the "neural Darwinist" model suggested by Gerald Edelman. I5 Knowing that such a model of the brain was correct would not, in and of itself, enable us to predict which inductions the person whose brain that was would make; that depends on the system(s) of hard-wired-in basic similarities, and (in the "neural Darwinist" model, on the operation of an analogue to natural selection in the unique individual brain) there may be a vast number of such systems (and selection events) at different levels of the brain's processing activity. Yet, if we verified that such a model was correct, we would hardly express the discovery by saying "the mind has turned out to be chaos". And the same thing goes if we discover that some model that does not come from computer science at all is the best model for the brain's activity. Many systems are too complex for us to survey and predict or simulate their activity in detail; this is not to say that we cannot seek useful theoretical models of such systems. To take an example from a totally different field, pessimism about the possibility of ever realistically simulating the behavior of an

RENEWING PHILOSOPHY

18 economy over a reasonably long period of time is not the same thing as pessimism about the possibility of a science of economICS. There is another side to Dennett's charge that I think the mind is chaos, however. Dennett is saying-and Fodor often says16-that pessimism about the power of computational models is scepticism about the possibility of "cognitive science". But the hidden premise in both thinkers' minds is a reductionist one. There is, in fact, an enormous amount of cognitive psychology that is not at all reductionist. There is no reason why the study of human cognition requires that we try to reduce cognition either to computations or to brain processes. We may very well succeed in discovering theoretical models of the brain which vastly increase our understanding of how the brain works without being of very much help to most areas of psychology, and in discovering better theoretical models in psychology (cognitive and otherwise) which are not of any particular help to brain science. The idea that the only understanding worthy of the name is reductionist understanding is a tired one, but evidently it has not lost its grip on our scientific culture.

2 Does Evolution Explain Representation? For the last three centuries a certain metaphysical picture suggested by Newtonian or Galilean physics has been repeatedly confused with physics itself. (More recently, metaphysical pictures suggested by biology and by computer science have been confused with those subjects themselves, in much the same way.) Philosophers who love that picture do not have very much incentive to point out the confusion-if a philosophical picture is taken to be the picture endorsed by science, then attacks on the picture will seem to be attacks on science, and few philosophers will wish to be seen as enemies of science. As far as our ways of understanding mind and language are concerned, the thrust of that picture was well captured by the claim of La Mettrie that man is a machine. The discovery of the idea of evolution by natural selection by Darwin and Wallace approximately a hundred years later seemed to add further evidence for the thesis that mind is to be understood by being reduced to physics and chemistry (we know from Darwin's journals that that is how he himself was inclined to see the matter). Today even materialist philosophers do not think that follows; it is on computer modeling, rather than on direct physical or chemical explanation, that thinkers of a reductionist bent, like my former self, now pin their hopes. But recently evolutionary theory has again come into play in

19

RENEWING PHILOSOPHY

20 discussions of the nature of mind, and of the relation of language to reality. Philosophers who apply the theory of evolution generally do so in a very simple way. The philosopher picks some capacity that human beings have, a capacity which it is in one way or another useful for human beings to have, and argues that it must have been selected for in the evolutionary process. This use of the theory of evolution is one that many evolutionary biologists find extremely questionable. 1 The working evolutionary biologist does not assume that every useful capacity of a species is the result of selection. A genetic alteration frequently has many different effects. If anyone of those effects contributes markedly to the reproductive success of members of the species having that gene, then that new genetic trait will be selected for, and other side effects, provided they are not so negative as to cancel out the benefits of having the new genetic trait, will be carried along. In this way, it can even happen that a trait which does not contribute to the survival potential or the reproductive success of a species, or even one which it would be better for the species not to have, arises through natural selection without itself being selected for. But it can also happen that the trait which is carried along is actually beneficial to the species, although that is not the reason why the trait became universal in the species. In general, the assumption that every change in a species which is beneficial to the species was specifically selected for is rejected in contemporary evolutionary theory. Evolutionists are extremely cautious about saying which capacities and organs and so on were specifically selected for (were "adaptations") in the evolutionary history of a species and which ones arose serendipitously. Philosophers, however, are not so cautious. 2

DOES EVOLUTION EXPLAIN REPRESENTATION?

21

Evolution, Language, and the World My primary concern in this chapter is with philosophical views of the mind, and with the way in which philosophical issues about the mind become entangled with various other issues. In a famous letter to Marcus Herz, Kant described the problem of how anything in the mind can be a urepresentation" of anything outside the mind as the most difficult riddle in philosophy. 3 Since the so-called linguistic turn in philosophy earlier in this century, that question has been replaced by the question uHow does language hook onto the world?" but the replacement has not made finding an answer any easier. Recently certain philosophers4 have suggested that the answer is provided by the theory of natural selection, either directly or indirectly. I want to examine this idea partly for its own intrinsic interest, and partly because it provides a natural transition to questions about the language-world relation. I will first state the idea in a very general way. Cognitive scientists have frequently suggested in recent years that the brain employs urepresentations" in processing data and guiding action. Even the simplest pattern:'recognizing devices in the brain could be described as producing representations. In his books Neurobiology and The Remembered Present, Gerald Edelman has described a neural architecture which could enable the brain to construct its own pattern-recognizing devices without uknowing in advance" exactly which patterns it will have to recognize. This architecture will enable a brain to develop a neural assembly which will fire whenever the letter A is presented in the visual field, for example, or alternatively to develop a neural assembly which will fire whenever the Hebrew letter aleph or a Chinese character is presented in the visual field, without having uinnate" A-recognizing devices or

RENEWING PHILOSOPHY

22 aleph-recognizing devices or Chinese-character-recognizing devices. If Edelman's model is right, then when an instance of the letter A is presented in my visual field and the appropriate neural assembly fires, that firing could be described as a ((representation" of the shape ((A". But the representations that neurobiologists, linguists, and computer scientists hypothesize go far beyond mere pattern recognizers. If an organism is to display what we call intelligence it is obviously useful, and perhaps necessary (as Edelman and others have suggested), for it to have, or to be able to construct, something that functions .as a map of its environment, with aspects that represent the various salient features of that environment, such as food, enemies, and places of refuge. At a higher level, such a map might be elaborated to show not only salient features of the current environment, but also a representation of the creature itself, and perhaps even a representation of the creature's psychological states C(self-consciousness"). In The Remembered Present, Edelman speculates about the sorts of neural architecture that might support a capacity to develop such representational schemata. All this is exciting science, or at least exciting scientific speculation. Whether or not it will payoff is not for me to judge, but of course I hope that it will. My doubt concerns whether neural science (or computer science, insofar as it leads to an increase in our ability to model the brain) can really speak to the philosophical question that I mentioned. What philosophers want to know is what representation is. They are concerned with discovering the ((nature" of representation. To discover that, in addition to the representations we are all acquainted with-thoughts, words, and sentences-there are other things which don't look like thoughts or words, things in the brain which it is useful to analogize to representations, is not to tell us what representation is. If a philosopher asks what the nature of representation is, and one tells him or her

DOES EVOLUTION EXPLAIN REPRESENTATION?

23

that there are tens of millions of representations in the Widener Library, one has not answered the question. And if one tells him or her that there are tens of millions of representations in human brains, one has not answered the question either. Or so it would seem. Let us take the form of the philosopher's question that I mentioned a few moments ago, "How does language hook onto the world?" Materialist philosophers generally favor one of two answers to this question. One kind of answer, which I shall not discuss here, uses notions from information theory. That answer has, however, run into apparently insuperable technical difficulties. 5 The other answer, which is today the favorite one among philosophical materialists, is that in the case of language, reference is a matter of "causal connection". The problem is to spell out the details, and in the next chapter I will examine one attempt to do this. Even before we look at such an attempt, it is apparent from the very beginning that there are going to -be difficulties with the details-whether those difficulties prove insuperable or not. One cannot simply say that the word "cat" refers to cats because the word is causally connected to cats, for the word "cat", or rather my way of using the word "cat", is causally connected to many things. It is true that I wouldn't be using the word "cat" as I do if there were no cats; my causal history, or the causal history of others from whom I learned the language, involved interactions with cats; but I also wouldn't be using the word "cat" as I do if many other things were different. My present use of the word "cat" has a great many causes, not just one. The use of the word "cat" is causally connected to cats, but it is also causally connected to the behavior of Anglo-Saxon tribes, for example. Just mentioning "causal connection" does not explain how one thing can be a representation of another thing, as Kant was already aware.

RENEWING PHILOSOPHY

24 For this reason, philosophers who offer this sort of account do not try to account for all forms of representation-that is too big a project to carry through at one fell swoop--but to account for forms of representation that might be thought basic, that is, for representation of observable objects in our immediate environment, such as trees and people and animals and tables and chairs. It is natural to suppose that the ability to represent such objects is the result of natural selection. My ability to understand the word "cat", for example, might involve my connecting that word with a more primitive representation of cats, a representation that is not itself a part of a language. It may be that in my representation of my environment, there is some "data structure" which "stands for" cats. To say that it stands for cats-this is the crucial move-is simply to say that an acccount of the evolutionary function of that data structure, and of the schematism to which that data structure belongs, will involve saying that that data structure enables us to survive and to reproduce our genes, or the entire schematism enables us to survive and reproduce our genes, because various parts of that schematism, including the data structure, have the "function" of corresponding to various things and kinds of things in the environment. Having the function of corresponding to things and kinds of things in that way just is representation, or rather, it is what we might call "primitive representation". If the problem of saying what representation is is not solved, the thought is, then at least progress has been made if this story is right. If this story is right, then we have at least a hope of saying what primitive representation is, and then philosophers can work on the task of showing how the more elaborate notion of representation that we actually possess is related to and grows out of primitive representation. Let me emphasize that, according to the view I am describ-

DOES EVOLUTION EXPLAIN REPRESENTATION?

25 ing, the intentional notion "stands for" can be defined by saying that "A stands for B" (where A is the data structure in the brain and B is the external referent) just means that "A stands to B in a relation (a correspondence, that is, a function from As to Bs) which plays such-and-such-a-role in an evolutionary explanation." "Stands for" is not being taken as primitive, if this works; "evolutionary explanation" is. The way in which I just explained the idea makes it sound as if the notion of a cat is supposed to be innate. While some thinkers-Jerrry Fodor is the best known-would indeed posit that the various data structures that make up the mental schematism that we use to represent our environment are indeed innate, others, like Gerald Edelman, would not. If Edelman's story is right, what is innate is not the mental representations themselves, but only the architecture which permits us to form such representations. While Edelman himself is wary of trying to answer the philosophical question about the nature of reference, someone who takes the line I have described could very well accept Edelman's model. What such a philosopher would have to do is insist that the architecture was selected for to perform the function of creating data structures which correspond in a certain way to objects in the environment and to kinds of objects in the environment. And again the claim would be that that correspondence, the correspondence which we have to talk about in explaining how the whole mental schematism came to be as the result of natural selection, is at least the primitive form of reference. 6 The idea is that natural selection is, so to speak, teleological: it produces things that have a telos or a ((function", and a structured way of performing that function. We can say what representation is by saying what the structures are that mediate representation, that is, how those structures function to enable the animal to survive and reproduce, and how a correspondence

RENEWING PHILOSOPHY

26 between structures in the head and things outside the head plays a role in that mediation, and then we can eliminate the mystery of teleology by showing how the teleology appears here, as it does, say, in the functioning of the foot or the functioning of the eye, as the result of familiar processes of natural selection. In what follows, I am going to confine attention to the form of the theory in which the individual representations are themselves innate, but it should be clear how the philosophical arguments should be modified if one takes the other form of the theory. To begin with, let us consider what it means to say that a trait was selected for. All explanations of selection involve counterfactual conditionals at one point or another. For example, if we say that the speed of the gazelle was selected for, what we will typically mean is something like this: that the gazelle needs to escape from certain predators, for instance from lions, if it is to survive and have offspring; and that the gazelles with a certain genotype, the one responsible for higher speed, lived to have offspring in greater numbers; and finally-we have to add-that they would not have survived in greater numbers if they had not run so fast: the lions would have caught them. This last addition is necessary, because without this addition the possibility is not ruled out that the genotype that is responsible for higher speed was selected for for some reason other than the higher speed that accompanies its presence.

If the gazelles had not run so fast, the lions would have caught them. This sentence, and others like it, "do the work" in powering explanations by natural selection. That an explanation by natural selection involves the use of counterfactuals is not a difficulty from the point of view of the philosophers I am talking about, since they are all committed to using counterfactuals, dispositional statements, and so on in the analysis of reference.

DOES EVOLUTION EXPLAIN REPRESENTATION?

27 There are, indeed, philosophers who regard counterfactuals as being just as problematic as reference itself-not to mention those, like Quine and Goodman, who regard counterfactuals as more problematic than reference itself-but this is an issue I have to defer to later chapters. The sense in which the gazelles' high speed is there for a purpose, the sense in which it has a Ufunction", is really rather minimal, which is why Ernst Mayr has proposed that we should speak not of teleology in evolutionary theory, but of teleologysimulation, or, as he puts it, Uteleonomy".7 Escaping lions is the function of the genetic trait in question only in the sense that if the high speed had not enabled the gazelles to escape the lions, then that feature would not have been selected for. Now, let us suppose that there are innate mental representations in the case of a certain species, say the dog. Let us suppose that one of the innate representations in the dog's brain is the representation of meat. What this will mean should be clear: we are saying that the dog's mental processes involve a Udata structure" which was selected for to do certain things; perhaps the data structure responds in a certain way when the dog sees meat, and this somehow triggers the appropriate responses, such as trying to get it, and eating it. Perhaps the data structure also operates in yet another way: when the dog wants meat, the data structure causes the dog to seek meat, or to whine to be fed, or whatever. Whatever the details may be, the point is that there are certain behaviors which involve meat and which involve the data structure, and the architecture which makes it possible for the data structure to mediate the behavior in that way was selected for. Again, any reference to teleology is unnecessary; all it means to say that this architecture was selected for this purpose is that if having a data structure which triggers these behaviors under these conditions had not

RENEWING PHILOSOPHY

28 enabled the dog to get food more often, then the dog would not have survived to reproduce its genes more often than other dogs with different genes. The important point (if some story like this one proves to be correct) is that the explanation of how the data structure came to be universal among dogs involves a certain "correspondence" between the data structure and meat.

Intentionality and Lower Animals One difficulty in evaluating the importance of these ideas is that we are all too ready to apply intentional predicates to lower animals in an unreflective way. We all say things like "the dog is trying to reach the meat", and we often seem to think that such descriptions mean that the dog has the propositional attitude of "thinking that what it sees is meat" just as they normally would if we were talking about a fully competent speaker of a natural language. Forget for the moment all evolutionary questions, and suppose we ask whether the dog really thinks that it is meat that it sees and reaches for, as opposed to stuff with a certain look and taste and edibiHty and so forth. Suppose we perform the following experiment: We make a "steak" out of textured vegetable protein. Let us suppose that our technology is so good that the lVP "steak" smells like real steak, looks like real steak, tastes like real steak, and so on. If one sees such a steak, one may well think, "I see a piece of meat". If one eats it, one may be perfectly happy. But if one is told that what one ate was textured vegetable protein, one will revise one's judgment, and decide that one didn't really eat a piece of meat, one ate a piece of TVP. (My oldest child, Erika, started distinguishing between "real" things and "make-believe" things-the beginning of the distinction between appearance and realityat about the age of two-and-a-half, by the way. I think that the

DOES EVOLUTION EXPLAIN REPRESENTATION?

29 appearance of this distinction between the "real" thing and the "unreal" thing is one of the most exciting developments of a child's language.) Now suppose that we give the synthetic steak to the dog. The dog eats the synthetic steak and is perfectly happy. Did the dog have a false belief? That is, did the dog believe that it saw real meat, just as we believed that we saw real meat, and not know that we had a false belief? Or did the dog's concept of meat include TVP "steaks" to begin with? The question makes no sense. A speaker of a language can decide that part of his or her concept of meat is that it should come from an animal, for example. A more sophisticated speaker can decide that it is part of the concept of meat that it should have the normal microstructure, whatever that may be. There is probably nothing in the dog's neural architecture which would allow it to think "this piece of meat came from an animal", and there is certainly nothing which would allow it to think "this piece of meat has a normal microstructure". To illustrate the same point in another way: Suppose we interpret the dog's concept, or as I would prefer to say, its "proto-concept", as referring not to meat but to whatever has a certain appearance and smell and taste. If the "meat" the dog ate on a certain occasion were not really a piece of meat, but a bit of ectoplasm which has been magically endowed with the right smell and taste and texture and appearance, the dog's thought that this is meat would be true of what it ate, on this interpretation, for its thought is not about meat in our sense, but only about the appropriate smell and taste and texture and appearance. Once again, a human being who discovered that what had just been eaten was not meat, and indeed not even a piece of matter, but a piece of ectoplasm, would take back the judgment that he or she had eaten meat. But the dog lacks the conceptual resources to make such a discovery. To deploy

RENEWING PHILOSOPHY

30

the jargon of philosophers of language, assuming dogs have proto-concepts, the dog's proto-concept of meat is "referentially indeterminate" in ways in which human concepts are not. Human concepts are less indeterminate because they enter into complex sentences, and human beings can say whether they believe or disbelieve those various sentences. In the case of the dog, those sentences are missing-sentences like "this meat has a normal molecular structure", "this meat came from an animal", "this meat is matter and not ectoplasm", and all the rest. But even the philosopher of language's way of putting the matter seems to me not radical enough. The real point is this: human beings are reflective creatures. Human beings are able to think about their own practice, and to criticize it from a number of points of view. If I have a thought and act on it, I can later ask whether my thought was successful or not, whether it achieved its goal, whether it contributed to my well-being, my satisfaction, and so on; but I can also ask whether my thought was true or not, and this is not the same question. I may decide that one of my thoughts was successful in enabling me to maximize my well-being, but was not in fact true. I was deceived, but the deception was a fortunate one. 8 No such cognitive performance is possible in the case of the dog. For a dog, the very distinction between having a true belief and having a successful belief simply does not make sense; and that means that the notion of a dog's thought as being true or false, and of its proto-concepts as referring or not referring to something, simply do not make sense. A dog can recognize that something is illusory only in the sense of encountering a disappointment. If something looks and smells like meat, but turns out to be made of rubber, then when the dog tries to chew it, it will experience disappointment. But the idea that even a successful encounter with "meat" may have succeeded although the belief was false is inapplicable in the

DOES EVOLUTION EXPLAIN REPRESENTATION?

31

case of a dog. Evolution didn't Udesign" dogs' ideas to be true or false, it designed them to be successful or unsuccesful.

Evolution Again With this ~~indeterminacy" of the dog's proto-concepts in mind, let us return to the evolutionary story. I postulated that if a certain data structure in the dog's brain (a proto-concept) didn't usually fire when the dog perceived meat, then dogs with a certain genotype wouldn't have survived to have offspring in greater frequency than did dogs with competing genotypes, as they in fact did. But the whole idea that a unique correspondence between the data structure and meat is involved in this bit of natural selection is an illusion, an artifact of the way we described the situation. We could just as well have said that the data structure was selected for because its action normally signals the presence of something which has a certain smell and taste and appearance and is edible. To this objection the defender of Uevolutionary intentionality" might reply that in fact the apparent indeterminacy if we look only at present-day dogs disappears if we consider evolutionary history. In the evolutionary history of the species, synthetic meat did not exist, for example. So, it might be argued, it would be wrong to regard the dog's proto-concept of meat as including synthetic meat. But it is difficult to see the force of this reply, since canned meat also didn't play any role in the evolutionary history of the dog, yet when a domestic dog sees some meat taken out of a can, the defender of evolutionary intentionality will presumably want to say that the dog thinks that that is meat, and that the dog's thought is true. It is also the case, by the way, that poisoned meat played no role in the selection process, since the dogs that ate poisoned meat did not survive to have offspring. Yet those who would take the dog's

RENEWING PHILOSOPHY

32

proto-concept to refer to meat would presumably say of the dog who sees and eats poisoned meat that it was right in thinking that what it saw was meat (although it didn't know the meat was poisoned), and not that what its proto-concept refers to is "unpoisoned meat". Yet, on the evolutionary story, why should one not say that the dog's concept of meat (and the human one too?) refers not to meat but to unpoisoned meat? Alternatively, why should one not just as well say that when the dog is given synthetic meat (or even poisoned meat) the dog thinks that that is ttmeat-stuff" (where the concept of meat-stuff is wide enough to include synthetic meat) and that the dog's thought is true; or why shouldn't one say that the dog's thought is "that's that· great stuff with such and such an appearance and such and such a taste", and that the dog's thought is true? Or, better, why shouldn't one just give up on talk of truth in connection with the thought of lower animals? Perhaps, if one wants to speculate, all that goes on is that certain ttmental phenomena" are associated with a feeling that a certain behavior is called for. Isn't it with dogs as with gazelles? Dogs which tended to eat meat rather than vegetables when both were available produced more offspring (gazelles which ran faster than lions escaped the lions and were thus able to produce more offspring). Just as we aren't tempted to say that gazelles have a proto-concept of running fast, so dogs don't have a proto-concept of meat. Indeed, in the case of the dog, there are a variety of different descriptions of the adaptive behavior: that certain dogs recognize meat better, or that certain dogs recognize food with a certain appearance and taste better, or just that certain dogs just recognize stuff with a certain appearance and taste better. The ttreference" we get out of this bit of hypothetical natural selection will be just the reference we put in our choice of a de-

DOES EVOLUTION EXPLAIN REPRESENTATION?

33 scription. Evolution won't give you more intentionality than you pack into it. 9

Reference and Counterfactuals The most telling argument against the idea that evolution explains intentionality is that the whole reference to evolution plays no real role in the "explanation" just sketched. What seems to give us an account is not the theory of evolution, but the use of counterfactuals and the appeal to considerations of selective reproduction of certain functions. But both of these strategies are available to any philosopher of language, and they require no particular reference to the theory of evolution . For example, a philosopher of language might very well take the following sorts of sentences as basic, at least in an initial investigation of reference: "I see an X", "I am touching an X", "I want an X", and so on. He or she might now say that when X is the name of a kind of observable thing, say a cat or dog, the way these sentences are "connected to the world" is the following: I would not normally assert that I see a cat, or I am touching a cat, or I want a cat unless I were (respectively) seeing a cat, or touching a cat, or wanting a cat. These claims are certainly correct. I wouldn't, in fact, normally assert that I see a cat unless I were seeing a cat, and so forth. Whether pointing to these counterfactuals is providing a sufficient explanation of what it is for the word "cat" to refer to cats is another question. But that question can be discussed whether or not the foregoing evolutionary speculations are true. In fact, as a biologist once remarked to me, people often forget that while biological evolution is Darwinian, cultural evolution is Lamarckian. What he meant, of course, is that in the case of cultural evolution we do see the inheritance of acquired characteristics, and there

RENEWING PHILOSOPHY

34 is no mystery about this. Suppose that, in fact, language is primarily the result of cultural evolution rather than of biological evolution, and that proto-concepts and so on play only a marginal role. The explanation of reference just suggested (using counterfactuals) would be no better and no worse off. If the idea is to give an account of intentionality by using counterfactuals, then we may as well discuss that idea directly, on its own merits, without the long detour through the at-present totally unproved speculations about our evolutionary history. 10 Of course, that evolutionary theory does not answer Kant's riddle as to how anything in the mind can be a representation of anything outside the mind does not mean that there is anything wrong with evolutionary theory, just as the fact that physics does not answer the riddle of the nature of free will and that brain science does not explain induction and language learning does not mean that there is anything wrong with physics or brain science. But I shall not belabor this point. I hope that these first two chapters have helped us to recall how different philosophical and scientific questions actually are, without denying that philosophy needs to be informed by the best available scientific knowledge. In the next chapter I shall look at an attempt by a well-known philosopher of cognitive science to solve Kant's problem-a philosopher who certainly appreciates that lower-level sciences will not, in and of themselves, solve Kant's problem (the puzzle of the existence of "intentionality", as it has come to be called), but who does think that it is possible to give an account, and who has put forward such an account for our consideration.

3 A Theory of Reference

In the preceding chapters I have tried to show that present-day science does not provide a sketch of an "absolute conception of the world", a sketch of a final metaphysics. In this chapter I will address an attempt by a well-known philosopher of cognitive science to solve Kant's problem, the problem of explaining the referential connection between our "representations" and the world. Jerry Fodor is one of the best-known philosophers of language working out of modern linguistic theory, especially the theories of Noam Chomsky, and he himself has been an important contributor to the field of psycholinguistics. Fodor's philosophy sometimes provokes violent disagreement (I disagree with much of it) but it is always enormously stimulating, and fertile in ideas. The new work is no exception. I find it of interest both intrinsically and also metaphilosophically-that is, from the point of view of what it reveals about contemporary philosophy. Fodor's new theoryl is rather complicated. I shall describe it in broad outline. For our purposes, we may consider the theory as beginning just where the discussion ended in the last chapter: Fodor examines an attempt to explain just what reference is using counterfactuals, points out what the problem is-and the problem is very closely related to the problems I discussed in the previous chapter-and then proposes a solution, namely

35

RENEWING PHILOSOPHY

36

the notion of Uasymmetric dependence". I need to introduce Fodor's terminology, however. Suppose someone makes an assertion which contains a token of the English word Ucat". In such a case, we shall say for short that the person has performed an act of 'cat' tokening". Cats cause Ucat" tokenings, but so do many other things; for example, I may see a picture of a cat and utter a sentence containing the word Ucat". In addition, I may utter the word Ucat" simply because I see the letters C-A-T, or because someone else asks me to repeat the word Ucat", but we shall exclude these purely syntactic causes of Ucat" tokenings from the discussion by fiat. We shall be interested in cases in which something extralinguistic causes a Ucat" tokening. Let us look a little more at the remark I just made, that is: U

(1) Cats cause ucat" tokenings. Fodor refers to this statement as a ulaw", b~t it is important to understand the notion of a law that he has in mind. Fodor is not thinking of laws in the sense in which Carnap thought of laws, that is, statements of fundamental physics which we can express as differential equations, or anything of that kind. 2 Fodor regards it as highly unrealistic (and of course I agree) to expect that linguistics should model its notion of law on the notion of law employed in fundamental physics. Linguistics, Fodor thinks, should be regarded as one of the uspecial sciences"3-like geology or evolutionary biology-which do not pretend to arrive at exceptionless universal generalizations. If geologists say that, other things being equal, rocks belonging to stratum A will always be found below rocks belonging to stratum B, Fodor will call this statement a law even though it contains the phrase Uother things being equal". Laws containing these Uother things being equal" clauses (Fodor calls them ceteris paribus clauses) are still able to support counterfactuals in many

A

THEORY OF REFERENCE

37

situations, and they enter into explanations which are fundamental in the particular special science, even if they are not fundamental in the sense that mathematical physics is fundamental. Of course, the statement that cats cause "cat" tokenings does not even mean that, other things being equal, a cat will cause a "cat" tokening. It rather means that cats frequently cause "cat" tokenings, or, perhaps, that they cause "cat" tokenings more often than anyone other kind of object causes "cat" tokenings. 4 But that is all right too, because all of the special sciences need to make statements about what is frequently the case, as well as statements about what is always the case, or what is always the case other things being equal. Now, the idea that Fodor is trying to work out is that what a word refers to is a matter of its causal "attachments to the world".5 At least in basic cases, what a word W refers to is going to be a matter of what causes W tokenings. But it is obvious that not everything that causes a W tokening is referred to by the word W. We already mentioned that pictures of cats cause "cat" tokenings, and the word "cat" does not refer to pictures of cats. Statues of cats and plastic cats may also cause "cat" tokenings. A meow may cause a "cat" tokening. And so on. The problem, then, is: given that there are many truths of the form (2) Xs cause "cat" tokenings, how can we determine which of these truths is fundamental, in the sense that it represents or determines the reference of the word "cat"? Note that Fodor's question is not: what does such-and-such a particular token of the word "cat" refer to? On a particular occasion, a token of the word "cat" may very well refer to a cat in a picture, or to a cat statue and not to a cat. Fodor's problem

RENEWING PHILOSOPHY

38 is to determine the "basic" meaning of the type word "cat". (What he wants to say about token reference, I do not know.) Fodor's answer is that there is a dependence relation among truths of the form (2), and that this dependence relation is asymmetric. The dependence is expressed by a counterfactual: (3) If cats didn't cause "cat" tokenings, then the other things that we mentioned (cat pictures, cat statues, the sound "meow", and so on) wouldn't cause "cat" tokenings either. In Fodor's terminology, the "law" (4) Pictures of cats cause "cat" tokenings

depends on the "law" that cats cause "cat" tokenings, but not vice versa, and it is this asymmetric dependence that determines the position of the law "cats cause 'cat' tokenings" at the top of the hierarchy of laws of the form (2). The fact that this law is at the top of the hierarchy is what makes it the case that the word "cat" refers to cats and not to pictures of cats, statues of cats, occurrences of the sound "meow", and so forth.

Is the Dependence Really Asymmetric? The first thing we have to consider is whether the dependence Fodor is talking about really exists and whether it is really asymmetric. The first question reduces to this: is it really true that if cats didn't cause "cat" tokenings, then pictures of cats wouldn't cause "cat" tokenings either? Fodor's thought is that if cats didn't cause "cat" tokenings then that would most likely be because the word "cat" didn't refer to cats. In the jargon of possible worlds semanticists,6 the idea is that the "closest possible worlds" (that is, the possible worlds which are closest to the actual world) in which cats don't cause "cat" tokenings are possible worlds in which the word "cat" refers to something else altogether. This seems reasonable

A

THEORY OF REFERENCE

39

(at least it seems reasonable if we take "possible worlds" to be hypothetical situations that are relevant to the truth value of a counterfactual, and not real worlds), so let us accept this for the sake of the argument. This shows that the dependence exists-that the ul aw" ttcat pictures cause tcat' tokenings" depends on the law Ucats cause tcat' tokenings"-but it doesn't suffice to show that the dependence is asymmetric. To show that the dependence is asymmetric, we have to show that it is not the case that if cat pictures didn't cause Ucat" tokenings, then cats wouldn't cause ttcat" tokenings either. Fodor takes this to be obvious, but is it? Wouldn't it be reasonable to suppose that the closest possible worlds in which it isn't a ulaw" that cat pictures cause Ucat" tokenings are possible worlds in which most people have no idea what cats look like? If we take those to be the closest possible worlds in which cat pictures don't cause ttcat" tokenings, then it would be the case that if cat pictures didn't cause ttcat" tokenings, then cats wouldn't cause Ucat" tokenings either, and the dependence would be symmetric. One possible counter would be for Fodor simply to insist on the Uintuition" that among those possible worlds in which cat pictures do not cause Ucat" tokenings, the closest ones are the ones in which people are blind, or can't recognize things in pictures, or otherwise have a nature which is very different from actual human nature, and not the ones in which people don't know what cats look like. But this seems implausible. A better move, I think, would be for Fodor to say that by Ucats cause tcat' tokenings" he means not that cats frequently cause Ucat" tokenings, but that they sometimes cause ttcat" tokenings. To this one might object that this claim is too weak; after all, isn't it the case that (5) For every observable kind of thing X, and for every word W, Xs sometimes cause W tokenings?

RENEWING PHILOSOPHY

40 One can think of some far-fetched situation, for example, in which the sight of an apple will cause me to mention dogs,7 or the sight of a pig will cause me to mention the moon. I think I know what Fodor would say about such cases. I think he would say that while it may be true that (6) Apples sometimes cause "cat" tokenings, that truth is not "lawlike". ("Lawlikeness" is a primitive notion in Fodor's metaphysics; moreover, it is a relation between universals and not a property of sentences, according to Fodor, which is how he meets the objection that using the notion of lawlikeness is employing a notion which is itself intentional.) While I have doubts about the supposedly non-intentional character of this notion, I shall, for the sake of argument, concede it to Fodor. Now, even if ordinary people had no idea what cats look like, if the word "cat" continued to refer to cats, there would presumably be at least some people (biologists and other specialists) who still knew what cats looked like. And therefore there would still be some cases in which cats still caused "cat" tokenings (unless we imagine a case in which the species has become extinct, which would perhaps raise still further problems for Fodor's theory). If the closest possible worlds in which ordinary people do not know what cats look like are still possible worlds in which some people do know what cats look like, then in those possible worlds it is still true that cats sometimes cause "cat" tokenings, and hence the existence of those "possible worlds" and their "closeness to the actual world" does not establish that the counterfactual if cat pictures didn't (sometimes)

cause "cat" tokenings, then cats wouldn't (sometimes) cause "cat" tokenings is true. It is still possible for Fodor to maintain that the dependence relation between the "laws" "cats cause 'cat' tokenings" and "pictures of cats cause 'cat' tokenings" is

A THEORY OF REFERENCE

41 asymmetric. But even if we assume, for the sake of argument, that Fodor's claims are correct, at least in the case of ordinary natural kind words like "cat", it does not necessarily follow that the theory does succeed in providing necessary and sufficient conditions for reference. In fact, the theory fails badly when the word in question is one whose extension is determined by an analytic necessary and sufficient definition. For example, let us suppose that someone introduces a word for someone whose wealth is at least a hundred billion dollars, say, "superbillionaire". Let us imagine that there are a small number, perhaps five or six, superbillionaires in the world, but that these people either have not heard of this neologism or despise it so much that they and their friends and families and close coworkers and bankers do not use it. Suppose that the half-dozen superbillionaires successfully conceal the fact that they are superbillionaires from the general public; the general public knows that they are billionaires but has no idea that they are superbillionaires. Then it could be that there is no single case in which a superbillionaire ever causes a "superbillionaire" tokening; yet it would still be true that "superbillionaire" refers to superbillionaires. Again there are a number of possible responses Fodor might make. He might, for example, say that if people knew all the relevant facts, then superbillionaires would cause "superbillionaire" tokenings. But what is a relevant fact depends on the meaning of the word we are considering. To know, that is, what the reference of "superbillionaire;" is, using the criterion that "if people knew all the relevant facts, then superbillionaires would cause 'superbillionaire' tokenings", we would have to know what facts are relevant to determining the truth value of such sentences as "X is a superbillionaire", and this would require having already interpreted "superbillionaire". Fodor might say that if English speakers were omniscient,

RENEWING PHILOSOPHY

42 then superbillionaires would cause "superbillionaire" tokenings. But omniscience is not only a non-actual state of affairs but an impossible state of affairs for human beings. Alternatively-and this is, I think, the most plausible line for Fodor to takeFodor might say that his theory is not meant to apply to words which have analytic definitions. 8 The trouble with this reply is that the whole raison d'etre of Fodor's theory is to be antihermeneutic; that is, Fodor rejects the view that one cannot determine the reference of a word in a language in isolation. According to hermeneuticists, whether of the Gadamerian variety or of the Davidsonian variety, to interpret a language one must make tentative assignments of extensions, that is, of reference, to the words; proceed to see whether the speakers of the language come out talking sense or nonsense according to these reference assignments; and then make adjustments in the tentative reference assignments until one finally ends up with an interpretation which makes maximum sense of the linguistic behavior of the people being interpreted. According to hermeneuticists there can be no such thing as necessary and sufficient conditions for a word W to refer to Xs. The best we can hope for are criteria of adequacy for translation schemes, or reference assignments, or assignments of truth-conditions. This view, the hermeneutic view, is anathema to Fodor. It leads, according to him, to "meaning holism", which in turn leads to "meaning nihilism",9 which leads to the denial of the possibility of a "special science" of linguistics. To determine whether a word in a language has an analytic definition or not, what we need is precisely an interpretation of the language. If Fodor's theory of reference applies only to a class of words, not to all the words of the language, and we can determine whether a word belongs to the class to which it applies only by first interpreting the language, then the theory does not do what Fodor wants it to

A

THEORY OF REFERENCE

43 do. It does not provide us with a reductive and non-holistic account of reference in terms of '~causal attachment". Fodor might reply that the theory is really meant to apply not to natural language but to his hypothetical innate language of thought, "Mentalese".l0 According to Fodor, all the concepts that appear in all natural languages are already available in Mentalese, available innately, and Fodor may well think that the innate structure of Mentalese determines which concepts have, and which do not have, analytic definitions. But if this is his view, then his entire theory is of no interest to those of us who find the idea that all concepts are innate preposterous. In any case, Fodor's theory fails on still other kinds of words as well. Consider the word "witch".ll It may be analytic that true witches have magical powers and that they are female; but having magical powers and being female is not a necessary and sufficient condition for being a witch. Many female saints are ascribed magical powers, but they are not considered witches. Nor must a witch have magical powers which come from an evil source or which are used for evil; anyone who has read The Wizard of Oz knows that there are good witches as well as evil witches. The word "witch" seems to have expanded its semantic range through a process of "family resemblance". The first witches to be mentioned in the Bible are, in fact, pagan witches; the characteristic Christian witch who has sold her soul to the Devil represents a much later idea. The problem posed by the word "witch" is that the "law" (6) Witches cause "witch" tokenings, is false. There are no witches to cause "witch" tokenings. I believe that Fodor would meet this objection by saying that still it is counterfactually true that

RENEWING PHILOSOPHY

44 (7) If there were witches, they would cause "witch" tokenlngs. But it is far from clear that this is true. If a witch must have magical powers, then it is far from clear that the concept of a witch is a coherent one, because it is far from clear that the concept of a magical power is a coherent one. We can certainly imagine possible worlds in which things regularly happen that superstitious people would regard as magic; but the very fact that they regularly happen in those possible worlds is strong reason for saying that in those possible worlds those things are not really magic-it is just that those worlds have different laws than the actual world. The notion of a world in which things happen that are "truly magical" is, I think, an incoherent one; and that means, I think, that the notion of a witch is an incoherent one. One might try to meet this difficulty by defining a witch not as someone who has magical powers but as someone who has supernatural powers, where the supernatural is understood not in terms of the notion of magic, but in terms of not falling within the categories of substance, space, and time. It is extremely doubtful that the pagan witches, or the witches of present-day African tribes, are supposed to derive their powers from something which is supernatural in that sense. It is a feature, in fact, of pagan thought that the gods, demons, and so on, are not supernatural in the sense which came into existence \\/ith the rise of Greek philosophy and the incorporation into the Jerusalem-based religions of a certain amount of Greek philosophy. The notion that what is magical must derive from the supernatural, in the philosophical/theological sense of "supernatural", is not part of the original meaning of the term.

A

THEORY OF REFERENCE

45 If the existence of witches is incoherent, then there are no possible worlds in which there are witches, and then (7) is senseless. But let us be charitable and suppose that somehow coherent sense can be made of the notion of a witch, and that there are possible worlds in which there are witches. Then, Fodor might say, the counterfactual (7), uif there were witches, they would cause 'witch' tokenings", is true. This, he might claim, is the ulaw" at the top of the relevant hierarchy generated by the asymmetric dependence relation. If this counterfactual is true-and I have just given reasons for thinking it isn't-then its truth is certainly not explained by natural law. For this counterfactual refers to what would be the case if some beings really had magical powers. If it is a truth that Uin the closest possible worlds in which there are beings of a certain kind with magical powers, those beings cause 'witch' tokenings among English speakers", then that truth is certainly not a truth which belongs to any natural science. It would, in fact, be a metaphysical truth. If Fodor's theory succeeded in this case, it would not provide a reduction of reference to the notions of the special sciences considered as natural sciences, but a reduction of the notion of reference to some very suspicious metaphysical notions. Similarly, if Fodor has to appeal to counterfactuals about what people would say if they were omniscient, he will again have to appeal to counterfactuals whose antecedent is physically impossible-impossible on the basis of natural law-and this involves the same kind of metaphysics. Now let us consider a perfectly ordinary word, usoldier". It is perfectly true that (8) Soldiers cause usoldier" tokenings. But it is also true that

RENEWING PHILOSOPHY

46 (9) People who pretend to be soldiers cause "soldier" tokenings. For the right "asymmetric dependence" to hold between these two ((laws", it is necessary that the following counterfactual be false: (10) If people who pretend to be soldiers didn't cause "soldier" tokenings, then soldiers wouldn't cause "soldier" tokenings. Now, of course, there is something funny about all of the counterfactuals that Fodor needs. For example, the counterfactual "if cats didn't cause 'cat' tokenings, then pictures of cats wouldn't cause 'cat' tokenings either" is not one that would ever be heard in ordinary life. If someone asked me "What would happen if cats didn't cause 'cat' tokenings?" my response would be to ask "What situation do you have in mind?". If the person said, "well, suppose the word 'cat' referred to a different kind of thing", then I might know how to evaluate various counterfactuals about what would happen if cats didn't cause "cat" tokenings. But to say that what I am to imagine is that the word "cat" has a different reference would be to beg the whole question, for Fodor's whole enterprise is to define reference without appeal to any notions which presuppose it. To do this, he must assume that counterfactuals like "if cats didn't cause 'cat' tokenings, then pictures of cats wouldn't cause 'cat' tokenings either" already have a truth value-that the semantics of counterfactuals (whatever that may come to) already assigns a truth value to such bizarre counterfactuals. If we are going to play that game, however, then I don't see why we shouldn't say that if people who pretend to be soldiers didn't cause "soldier" tokenings, that would almost certainly be because the word "soldier" had a different reference. After all, wouldn't we expect that as

A

THEORY OF REFERENCE

47 long as the word usoldier" continues to mean soldier, then people who tell other people that they are soldiers cause usoldier" tokenings? If the closest possible worlds in which people who pretend to be soldiers do not cause usoldier" tokenings are the worlds in which usoldier" has a different reference, however, then the counterfactual Uif people who pretend to be soldiers didn't cause 'soldier' tokenings, then soldiers wouldn't cause 'soldier' tokenings" is true, and not false as the asymmetric dependence theory reqUIres. Fodor would perhaps say that if people were able to distinguish pretend soldiers from real soldiers infallibly, then soldiers would cause usoldier" tokenings, but people pretending to be soldiers would not cause usoldier" tokenings; but worlds in which people have such extraordinary abilities would seem to be extremely far from the actual world. Moreover, the notion of knowledge which is involved in the description of such possible worlds is itself one which presupposes the possession of a language, that is, the ability to refer.

Fodor's Notion of a Cause In this discussion I have not questioned Fodor's free use of the notion of something causing something else, e.g., a cat causing an event which is a Ucat" tokening. But let us examine this notion a little more closely. The notion of cause in ordinary language is both context bound and interest dependent, as Hart and Honore pointed out many years ago. 12 For example, if John eats foods high in cholesterol for many years and refuses to exercise, against the advice of his doctor and even though he has been told he has high blood pressure, and as a result suffers a heart attack, we may say that (i) his failure to exercise and eat a proper diet caused the heart attack, or that (ii) his high

RENEWING PHILOSOPHY

48 blood pressure caused the heart attack, depending on the context and our interests. If fact, even if A is a contributory cause of B we are unlikely to refer to A as "a cause" of B unless A is the sort of contributory cause that it would be natural to refer to as "the cause" of B in at least some contexts. For example, if a pressure cooker has a stuck valve and explodes, the absence of holes in the vessel of the pressure cooker is clearly a contributory cause of the explosion, as any engineer will tell you, but we would never in ordinary language say that the absence of holes in the vessel of the pressure cooker was "the cause" of the explosion, nor would we normally say

(II) The absence of holes in the vessel of the pressure cooker was a cause of the explosion, although we would say that the stuck valve caused the explosion (and to say that a valve is stuck just means that there isn't a hole in a certain place where there ought to be a hole). The notion of cause that Fodor is using is just this ordinarylanguage, context-sensitive, interest-relative notion, and not the relatively more context-independent notion of contributory cause. To see this, consider what would happen if we tried to interpret Fodor's theory using the notion of contributory cause as the relevant notion. When Fodor says "cats cause 'cat' tokenings", what he means on this interpretation is that the presence of a cat is a contributory cause of many cat tokenings. But then it is certainly true that the past behavior of English speakers (not to mention speakers of Anglo-Saxon and other ancestors of English) is also a contributory cause of "cat" tokenings, since we would not be using the word at all if the past linguistic behavior of English speakers (Anglo-Saxon speakers, etc.) had been different. So

A

THEORY OF REFERENCE

49 (12) The past behavior of English speakers causes "cat" tokenings would also be true on this interpretation of causes; that is, the past linguistic behavior of English speakers is a contributory cause of "cat" tokenings. What is the relation of counterfactual dependence between (12) and "cats cause 'cat' tokenings"? Is the dependence asymmetrical? And if so, in which direction is the asymmetry? Well, if cats didn't cause "cat" tokenings, the word "cat" would probably mean something else (this assumption is essential to Fodor's whole argument), but even if it meant something else, as long as it was still a word in English, it would still be the case that the past behavior of English speakers was a contributory cause of present "cat" tokenings. It is not true that (13) If cats didn't cause "cat" tokenings, then the past behavior of English speakers wouldn't cause (be a contributory cause of) "cat" tokenings. On the other hand, if the past behavior of English speakers were not a contributory cause of "cat" tokenings, that would almost certainly have to be because "cat" wasn't a word in the language. Thus it is reasonable to say that the set of closest possible worlds in which the past behavior of English speakers is not a contributory cause of "cat" tokenings is the set of possible worlds in which the word "cat" is not a word in the English language. If so, then the following counterfactual is true: (14) If the past behavior of English speakers didn't cause (wasn't a contributory cause of) "cat" tokenings, then cats wouldn't cause "cat" tokenings. On this interpretation of "cause" as "contributory cause", (i) the dependence is asymmetrical, and (ii) the dependence goes the wrong way for Fodor's theory.

RENEWING PHILOSOPHY

50

However, this is, as I said, pretty clearly not the interpretation of cause that Fodor has in mind. All of his examples, as well as his references to other special sciences like geology, indicate that he simply wishes to take the ordinary-language notion of cause as primitive. What is strange about this, is that this notion of causation is interest relative. Whether we say that A caused B or not depends on what we take to be the relevant alternatives. If we are interested in what would have happened to John if he had obeyed the doctor's orders, then we are likely to say that his eating habits and lack of exercise caused his heart attack, but if we are interested in what would have happened to John if he had not had high blood pressure to begin with, then we are likely to say that his high blood pressure caused the heart attack. Notice that being int~rested in something involves, albeit in a slightly hidden way, the notion of "aboutness", that is, the central intentional notion. To be interested in something, in this sense, you have to be able to think about it-you have to be able to refer to it, in thought or in language. Fodor uses a notion which has an intentional dimension; his notion of things "causing" other things is not a notion which is simply handed to us by physics. For in fundamental physics, at least, one usually ignores the distinction between contributory causes and "the cause", and tries to provide a formalism which shows how all of the factors interact to produce the final result. Moreover, Fodor assumes that counterfactuals have definite truth values, including many counterfactuals that would baffle any ordinary speaker. Here his defense is that counterfactuals are used in geology, evolutionary biology, and other "special sciences". He might also say that I have made his use of counterfactuals look more suspect than it is by referring to the possible-worlds semantics for counterfactuals. Perhaps Fodor would reject the possible-worlds semantics. He could certainly say that, whatever the status of that semantics, all he is assuming

A

THEORY OF REFERENCE

51 is that counterfactuals make sense and have truth values, not that it makes sense to talk about possible worlds, or about possible worlds being closer to or farther away from the ~ctual world. This would certainly be fair; and I did not mean, in using this vocabulary, to indicate that I myself regard possibleworlds talk as metaphysically satisfactory, or that I regard talk of closeness of possible worlds as explaining how it is that counterfactuals are true. The possible-worlds semantics, as I see it, is simply a model which enables us to formalize the deductive relations among counterfactuals, and no more. But that formalism does have a close connection with counterfactuals in the following way: just looking at how we employ counterfactuals, we will notice that-even if we think of possible worlds as mere stories or hypothetical situations-we do not treat all possible worlds in which the antecedent is true as equally relevant to the truth of a counterfactual. If I say "If I had put this lump of sugar in my coffee, it would have dissolved", then there are an enormous number of possible worlds in which the antecedent of that counterfactual is true-that is, there are an enormous number of hypothetical situations in which the antecedent is true-but I do not regard all of them as relevant. In the case of that counterfactual, for example, I would ignore hypothetical situations in which the laws of physics are different. From some counterfactuals that exclusion, of hypothetical situations in which the laws of physics are not as they are in the actual world, suffices. If A and B are related in such a way that the material conditional "if A, then B" follows deductively from a finite set of true laws of physics, then the counterfactual "if A were the case, then B would be the case" is certainly true. For counterfactuals of this kind-I have elsewhere called them "strict counterfactuals"-a ((similarity metric on possible worlds" is easy to specify: a world is "sufficiently close" to the actual world just in case it obeys the same laws of

RENEWING PHILOSOPHY

52 physics as are true in the actual world. But many counterfactuals are considered as true even though there exists some physically possible world in which the antecedent is true and the consequent is false. For example, there are physically possible worlds in which a lump of sugar is put in hot coffee and does not dissolve. This is true because of entropy considerations and also because of quantum mechanical considerations. 13 We regard these possible worlds as somehow too abnormal to falsify the counterfactual "if I had put the sugar in the coffee it would have dissolved". So even if we drop the language of "possible worlds", and drop the language of possible words being "closer to" or "farther away from" the actual world, we still have to recognize that a counterfactual may be true even though there exists a hypothetical situation (one representable as a physical state of affairs in the formalism of physics) in which the antecedent of the counterfactual is true, all the fundamental physicallaws which hold in the actual world hold in the hypothetical situation, and the consequent of the counterfactual is false. Indeed, almost all of the counterfactuals that we use in the special sciences are of this kind. For this reason, we may say that the very use of counterfactuals, whether it explicitly presupposes a similarity metric on possible worlds or not, implicitly determines such a metric. By looking at which counterfactuals we count as true, we can see which hypothetical situations we consider more relevant to determining the truth value of the counterfactual and which we consider irrelevant (or abnormal); and if we call the more relevant ones "closer", then it does not seem that any great harm will be done. Still, Fodor might reply, this way of looking at the situation is misleading because it gives an unreasonable pride of place to the laws of physics, that is, it gives too much weight to what is physically possible and what is physically impossible, and not enough weight to "laws" in another sense, the laws of the special

A

THEORY OF REFERENCE

53 sciences. Why should we, after all, emphasize possible worlds which can be represented as states of affairs in the physicist's sense-regions in phase space, ·states of affairs whose timedevelopment obeys the laws of fundamental physics, possible worlds in which nothing "physically impossible" happens? If we think of such states of affairs as unproblematic, and then think inherently problematic any selection from those states of affairs which involves treating some of them as more relevant than others, then, of course, we will be led to look for something like a "similarity metric on possible worlds" in order to account for the fact that there are true counterfactuals which are not what I called "strict counterfactuals". (Another motive for looking at possible-worlds semantics, a very different one, is to have some account of the truth of some counterfactuals whose antecedents are physically impossible.) To meet this objection, suppose we took the view that the laws of the special sciences are just as unproblematic as the laws of physics, and that it is at least a sufficient condition for the truth of a counterfactual that (i) there are situations in which the antecedent holds and in which all laws of nature (not just the laws of fundamental physics) hold; and (ii) the consequent holds in all those states of affairs. This suggestion doesn't quite work (which may be why Fodor does not put it forward). To see that it does not work, consider again the counterfactual "if I had put this lump of sugar in my coffee, it would have dissolved". According to the suggestion being considered, that counterfactual is true if the lump of sugar dissolves in all hypothetical situations in which it is put into the coffee and all the laws of the special sciences, including laws with an "other things being equal" clause in them, hold. Now, the suggestion may seem to work, because one law of ordinary chemistrywhich is certainly a "special science"-is "other things being

RENEWING PHILOSOPHY

54 equal, a piece of sugar will dissolve in hot coffee". But things are not so simple. For the truth of the statement that other things being equal a piece of sugar will dissolve if put into hot coffee is perfectly compatible with its being the case that this piece of sugar will not dissolve if put into hot coffee. It could be the case that this piece of sugar is in some way "not normal". The truth of a law of the form "other things being equal, an A will do B in circumstances C" does not imply that any particular A will do B in circumstances C, nor does it imply any counterfactual of the form "this particular A would have done B if it had been in circumstances C". This is why I think that Fodor doesn't want one to try to give necessary and sufficient conditions for the truth of a counterfactual using any notion of "law", not even his broadened notion. Here is what I myself think about counterfactuals: I think that what makes a counterfactual true is simply that the consequent follows from the antecedent together with various relevant natural laws or general truths, plus the initial and boundary conditions in those situations that it wOllld be reasonable to regard as compatible with the intentions of the speaker who uttered the counterfactual. By this I don't mean to suppose that a speaker who utters a counterfactual can imagine all conditions that might be relevant to what the speaker has in mind. To describe the conditions which are relevant to what he or she has in mind might require quantum mechanics, for example, or even some physical theory which has not yet been thought of, or some theory in some other science which has not yet been thought of. What I am saying is that an evaluator of the counterfactual who does know the relevant theories, and who thinks of some case in which the antecedent of the counterfactual is true and the consequent of the counterfactual is false, must decide whether the case actually falsifies the counterfactual, or whether it is more reasonable to think that the case is

A

THEORY OF REFERENCE

55 too far-fetched to be relevant to what the speaker intended to say. In effect, the evaluator imagines himself as being the speaker, and as having the same intentions as the original speaker, but with much more knowledge of natural science, or whatever, than the original speaker. Fodor would presumably reject this suggestion because it implies that counterfactuals ineliminably presuppose the point of view of reason. To say that taking into account one state of affairs is "reasonable" while taking another state of affairs into account would be "unreasonable" is to make a normative judgment; and what I am going to argue in the lectures to come is that the desire to leave all normative considerations out of the philosophy of language is precisely what leads to the failure of the various attempts to "naturalize the intentional".

Fodor's Metaphysical Picture Let me now say a word or two flbout Fodor's metaphilosophical stance. It is clear that Fodor's aim is to provide a reductive account of reference in the following sense: he views the special sciences which do not especially refer to the human race, sciences like evolutionary biology and geology, as somehow belonging to the realm of the non-intentional (or the preintentional). 14 These sciences describe reality as it is, in itself, before symbol users and referers appear on the scene. Moreover, he seems to think 15 that any concept which is essential to such a special science is philosophically kosher, that is, is a concept which we have a philosophical right to regard as unproblematic. 16 (This is somewhat strange, since one would think offhand that what we have a right to regard as unproblematic should depend on the question being asked.) Geology uses laws with "other things being equal" clauses in them; therefore, "other things being equal" is an unproblematic notion which we can

RENEWING PHILOSOPHY

56

use anywhere in philosophy. Geology uses the ordinary notion of something causing something else; therefore the ordinary notion of causation is an unproblematic notion which we can use anywhere in philosophy. Geology uses counterfactual conditionals, therefore the counterfactual conditional is an unproblematic notion which we can use anywhere in philosophy. I certainly agree that there is nothing wrong with a geologist, or an evolutionary biologist, or a psychologist, or whoever, using a counterfactual conditional, or saying that something is the cause of something else, for that matter. My opposition to Fodor is by no means diametrical. Fodor's view has both a reductionist and an anti-reductionist element. Fodor is antireductionist in that he does not think that the special sciences can be reduced to fundamental physics. He has argued for many years and very persuasively that it is impossible to require geology or evolutionary biology or psychology to use only concepts which are definable in terms of the concepts we need to do mathematical physics. This anti-reductionist side of Fodor's thought is one that I hail. Fodor also recognizes that the nature of reference and the nature of general intelligence are not questions to which present-day Artificial Intelligence offers even a sketch of an answer. He would have no quarrel, as far as I can see, with the arguments I advanced in the first chapter. At the same time, however, Fodor clings to the picture of the natural sciences-that is, the sciences below psychology-as describing the world as it is independent of intentional or cognitive notions, a ready-made world. His anti-reductionism simply leads him to view any notion used by those sciences as descriptive of what is "out there" independent of mind and independent of language users. If geology needs counterfactual conditionals, then counterfactual conditionals are "out there"; if geology needs to say that something causes something else, then causation (in that sense) is "out there". His retention of

A

THEORY OF REFERENCE

57 the old desire to reduce consciousness and mind to something wholly physical remains, but in an altered form, as the desire to reduce the central intentional notion, the notion of reference, to these notions. It is this reductionist side of Fodor's thought that I reject. But I have argued here that even if I give Fodor all that he wants to use-his notion of "law", his peculiar counterfactual conditionals, his notion of causation-still he has not successfully defined reference in such terms. The reduction has failed. On Fodor's picture, just as physicists discovered at a certain stage of the game that you cannot give a complete description . of the world in terms of the forces known to classical physics alone, and so had to add new primitives in order to talk about new forces-for example the so-called "strong force" that holds the nucleus of the atom together-so geologists and biologists discovered that, just as the strong force and the weak force and mass and charge are out there in the world, independent of mind, so "ceteris-paribus-hood" (or "other-things-being-equalness") is out there in the world, independent of mind, and one thing's "causing" another is out there in the world, independent of mind, and something's being a ttlaw" is out there in the world, independent of mind. It is at this point that Fodor and I part company. From the fact that a statement is not explicitly about anything mental it does not follow that none of its presuppositions make any reference to our cognitive interests, our way of regarding different contexts, or our intentional powers. I have argued that the notion of causation, for example, has a cognitive dimension, even when we use it in a statement about inanimate objects, for example the statement that the stuck valve caused the pressure cooker to explode. The cognitive or ttintentional" dimension lies in part in the presupposition that hearers of the statement regard such facts as that the vessel of the pressure cooker does not have a hole in it as ttbackground

RENEWING PHILOSOPHY

58 conditions" which may be taken for granted, as well as in our knowledge of the salience that others attach to the condition of the valve. When I say that the water would have boiled if the gas stove had been turned on, there is an even more complex cognitive dimension. If Fodor (the side of Fodor that I agree with) is right, and ordinary-language notions of causation (as opposed to the tech- . nical notions used in mathematical physics), and ordinarylanguage counterfactual conditionals, and ordinary-language notions of scientific law (however much Fodor may overblow and distort both counterfactual conditionals and the ordinary notion of scientific law) are essential in sciences which do not deal especially with human beings (like geology or evolutionary biology), what that shows is not that the statement that one thing caused another has no intentional dimension, but rather that concepts which do have an intentional dimension, concepts whose very use presupposes an identification with the interests and saliencies of human beings rather than a "view froITI nowhere", are indispensable even when we talk about rocks or species. To deny, as I do, that there is a "ready-made world" is not to say that we make up the world. I am not denying that there are geological facts which we did not make up. But I have long argued that to ask which facts are mind independent in the sense that nothing about them reflects our conceptual choices and which facts are "contributed by us" is to commit a "fallacy of division". What we say about the world reflects our conceptual choices and our interests, but its truth or falsity is not simply determined by our conceptual choices and our interests. To try to divide the world into a part that is independent of us and a part that is contributed by us is an old temptation, but giving in to it leads to disaster every time. If one accepts this point of view, then both the successes of Fodor's criticism of

A

THEORY OF REFERENCE

59

reductionism with respect to the special sciences and the failures of his own attempt at reductionism with respect to semantics take on a different aspect. It does not look as if the intentional can simply be reduced to the non-intentional; rather, it begins to look as if the intentional intrudes even into our description of the non-intentional, as if the intentional (or, better, the cognitive) is to some extent ubiquitous. One thing that interests me in this book is why we are so reluctant to admit this. What does it show about our culture and our entire way of thinking that it is so hard for us to admit this, and what might a philosophy might be like that began to give up all reductionist dreams?

4 Materialism and Relativism

While individual philosophers continue to produce and defend as wide a range of metaphysical views as they ever have, today two outlooks have become dominant in American and French philosophy; these are, respectively, the outlooks of materialism and of relativism. Although few American philosophers actually call themselves materialists, and I do not know of any French philosophers who actually call themselves relativists, the terms "physicalism" and "naturalism" have become virtually synonymous with materialism in analytic philosophy, while the thrust of deconstructionist views is often clearly relativist, if not downright nihilist. I have argued for some years that both styles of thought are too simplistic to be much help in philosophical reflection. 1 I have already indicated some reasons for my dissatisfaction with materialism as an ideology. The materialist philosopher believes that present-day scientific theories already contain the broad outlines of a solution to philosophical problems about the nature of minds and intentionality; I have argued that there is no reason to believe that this is the case. However, my purpose here is not to polemicize against materialism and relativism, but to see what we can learn from the failure of these large points of view. There are, happily, many philosophers who reject both relativism and semantic materialism, that is,

60